Why it matters
Without AI, data interpretation relies on manual review, risking missed buried features, inconsistent time slices, and untracked changes that slow build, complicate reporting, and hinder post-excavation analyses and archiving.
From design to safety
Design
AI-driven survey design
AI analyzes historical surveys, site context, and sensor specs to propose GPR survey lines, spacing, and depth targets; it checks alignment with the written scheme of investigation and anticipated features.
- GPR antennas
- time slices
- data processing
Build
Real-time data acquisition checks
AI monitors incoming traces, flags noise and clutter, and recommends on-site reconfiguration of antenna polarization and line spacing to optimize coverage.
- GPR antennas
- time slices
- data processing
Operation
Feature detection and mapping
AI fuses time slices with depth estimates to produce feature maps, highlights reflections corresponding to buried walls, and updates context sheets and strip maps.
- time slices
- depth estimates
- reflections
- buried walls
Maintenance
Data integrity and calibration
AI compares new scans against baseline, recalibrates depth estimates, and logs processing parameters for post-excavation assessment and archive deposition.
- depth estimates
- data processing
- reflections
Safety
Quality assurance for safety limits
AI flags high-risk trench zones by corroborating buried services and UXO indicators, guiding site-setup adjustments and safety briefings with the supervising archaeologist.
- buried walls
- data processing
- reflections
- time slices
Part by part
- 1
GPR antennas
- Watch for
- amplitude drift, phase shift, and coupling effects across adjacent scans indicating sensor misalignment or soil heterogeneity
- AI action
- detect anomalies in signal strength and trigger recalibration of antenna orientation and gain
- 2
time slices
- Watch for
- inconsistent slice thickness or alignment gaps that degrade 3D interpretation
- AI action
- flag discontinuities and prompt re-run of limited line scans for coverage
- 3
depth estimates
- Watch for
- depth variance beyond expected lithology changes or layer interfaces indicating calibration drift
- AI action
- adjust velocity model and notify field supervisor of potential re-sampling
- 4
reflections
- Watch for
- unexpected strong reflections near known targets or at depth suggesting boundary anomalies
- AI action
- classify as probable edge of feature and add to feature map with confidence score
- 5
buried walls
- Watch for
- linear high-amplitude reflections with coherent geometry indicating wall edges
- AI action
- trigger trenching plan update to expose feature safely
- 6
data processing
- Watch for
- processing pipeline failures or mismatches between raw traces and migrated results
- AI action
- pause processing, switch to validated workflow, and log processing provenance
Worked example
Catching a buried wall before trenching
Illustrative scenario with sample data.
A development-led site of 5 hectares with Bronze Age to Iron Age deposits requires rapid identification of buried walls to inform trench plans and avoid unnecessary disturbance.
Signal in
siteAreaHa: 5, periodsFound: Bronze Age to Iron Age, sensor: multi-channel GPR, dataRate: high, baselineProcessed: true, depthTarget: 2.5 m, anticipatedFeatures: buried walls
System out
AI-generated feature map highlights buried walls with confidence 0.78, depth range 1.2–2.8 m, recommended trench plan along wall projection, and updated written scheme of investigation references.
Human sign-off: Project manager reviews AI output, confirms trench alignment with the witness to avoid disturbance, and signs off for immediate trenching adjustment.
How to roll it out
Phase 1
Weeks 1 to 4: Baseline
Establish data pipelines, configure hardware, and calibrate GPR arrays with joint QA checks and initial feature cataloging.
Phase 2
Weeks 5 to 12: Pilot scanning
Run targeted surveys across test zones, validate AI outputs against known features, and refine velocity models and interpretation workflows.
Phase 3
Weeks 13 onward: Full integration
Scale to the full site, integrate with HER, and produce continuous reporting through context sheets, plans, and daily field summaries.
Where people stay in charge
- Decisions to suspend or modify survey design must stay with the project manager and lead geophysicist, who verify safety and site constraints.
- The qualified person overseeing heritage consent and HER updates must approve any changes to the written scheme of investigation or feature interpretation before work proceeds.
- All archival deposition and post-excavation assessment decisions require sign-off by the curator or heritage consultant before archiving.
Questions owners and crews ask
How does AI improve GPR survey planning for a complex site?
AI analyzes site history, expected stratigraphy, and sensor characteristics to propose optimized line spacing, orientation, and depth targets, reducing ambiguity in early interpretation.
What happens if the AI detects a potential buried featurelate in the day?
The system flags the feature, updates the feature map, notifies the site supervisor, and triggers a safety brief and controlled trench planning adjustment.
Can AI help with documenting context sheets and plan updates?
Yes, it generates linked contexts and evolves the Harris matrix, ensuring changes are traceable and compatible with single context recording standards.
Who signs off on AI-generated safety decisions for buried services?
The site safety officer in collaboration with the project manager reviews and signs off, ensuring compliance with regulatory and planning conditions.